One of the question marks hanging over solid state drives (SSDs) is their durability. Through decades of experience, we know hard drives have mean times before failure in the millions of hours, but with commercial SSDs being less than a decade old, it's not as clear.
Well, The Tech Report took up the challenge of SSD durability and made an impressive finding. As part of its durability test, it forced six drives from Intel, Kingston, Samsun and Corsair to continuously write and rewrite 10GB of static data consisting of files of various sizes. This test began in August 2013 and ran almost a year of constant writing to the disks.
The results were nothing short of remarkable. The Intel 335 Series is designed to take its own life after a pre-determined volume of writes, and Tech Report's drive committed seppuku after writing around 750TB of data, and that was after multiple SMART warnings were issued.
SMART, or Self-Monitoring Analysis and Reporting Technology System, is a firmware chip on hard disks and SSDs that continuously monitors the drive for things like performance issues and error rates. It has a technology known as predictive failure analysis that warns you when a hard drive failure is imminent. So it does not prevent hard drive failures, but it does warn you when failure is imminent.
The Kingston HyperX 3K made it to 728TB before it died, and it had reallocated nearly a thousand sectors before it expired. Again, TR says the wear indicator and SMART warnings provided plenty of notice that the drive was ready to die off.
Corsair's Neutron GTX made it up to 1.1PB, but then it fell apart rapidly at around 1.2PB of data. The Samsung 840 Series started reporting reallocated sectors very early, after just 100TB, but they attributed that to the Samsung NAND being more sensitive to voltage-window shrinkage than the MLC flash in the other SSDs. The 840 Series made it past 900TB, but its reliability was compromised well before that.
The Samsung 840 Pro, a high-end version of the regular 840, and a second Corsair HyperX 3K were able to make it to 2PB without issue.
Now to put this in perspective, your C: drive writes an average of 20GB to 40GB on an annual basis, depending on where you store your data. If like me you have a large D: hard disk for data, then that offloads the amount of I/O on the C: drive. Most disk writes over the course of the day are swap file data.
So writing 2PB basically means you can get 1,000 years out of a SSD. And that's the commercial grade SSDs. The ones used in enterprise servers from EMC, IBM, HP, etc, are even higher grade than that. These drives usually deal with wear by packing more storage than they are rated. A 512GB drive might have 600GB, for instance, but only 512GB accessible. As cells wear out and are marked as bad, the inaccessible memory begins to take over for the bad cells.
Whatever tricks are used, it's an interesting test. The fact it took a year of writing data non-stop is remarkable. No doubt there will be more tests like this and if they come out similar, it should give people more confidence in SSD drives. If you want to check on the health of your SSD, try SSD Life, which offers a free trial.
